* Resolved an issue with the ESP8266 where gamma correction would be
performed twice. Changed GAMMA_CORRECTION to SOFTWARE_GAMMA_CORRECTION
to make a distinction between software and firmware gamma correction.
The ESP8266 does firmware gamma correction and dithering, while the
Raspberry Pi uses slightly more inferior software gamma correction.
Changed the software gamma table to match the gamma table used in the
ESP8266 firmware.
* Improved the spectrum visualization by using one of the color channels
to visualize the absolute value of the temporal derivative of the
spectrum. Also added a feature to reject the "common mode" spectral
components, which is analogous to the spectral DC component.
* Signficantly improved the GUI and added a frequency adjustment slider.
Adjusting the frequency range has a big impact on the visualization
output. Recommend using a high frequency range (something like 4 kHz -
10 kHz) when running the scrol visualization.
This commit adds support for the Raspberry Pi, which allows users to
create a completely standalone music visualization system. The Raspberry
Pi should be connected directly to a ws2812b LED strip. A PWM-capable
GPIO pin should be connected to the data line of the LED strip. A USB
audio input device should be connected to one of the Raspberry Pi's USB
ports.
It is recommended that the GUI and FPS output be disabled when running
the visualization on the Raspberry Pi. These features can degrade
performance on the already computationally limited Raspberry Pi.
Fixed a bug where certain config.FPS values would result in arrays with
mismatched dimensions. The mismatched dimensions occurred because an
integer was rounded instead of truncated. This is now fixed and
exceptions should no longer be raised when certain FPS values are used.
Fixed a bug where the GUI ComboBox value would sometimes not match the
effect being displayed. This bug sometimes occurs when the GUI is first
initialized. This change now ensures that the ComboBox value is
correctly initialized.
Added a ComboBox to the GUI to allow the user to change the
visualization effect without having to restart Python. The user can now
select from one of the following visualizations using the GUI: 'Scroll
effect', 'Spectrum effect', 'Energy effect'
Fixed an error that caused strange byte encoding in Python 3. This error
was caused by a mistake in porting code from Python 2.7 to Python 3 and
would cause strange LED flickering behaviour. The LED update code should
now work properly in Python 2.7 and Python 3.
Fixed some bugs that occurred in Python 3.5 but were not present in
Python 2.7. Most compatiblity issues were caused by incompatible type
casting of numpy arrays.
The Mel filterbank visualization is relatively simple but produces
visualizations that are quite nice. Run this file to view the mel
filterbank visualization
Changes to config.py:
> Increased STFT rolling window size
> Added configurable options for setting the max and min FFT frequency
Changes to dsp.py:
> Significantly improved the normalized weighted phase deviation code.
This noticeably improves the onset detection accuracy.
> Logarithmic partitioning of the FFT bins now happens after onset
detection instead of before onset detection. This improves onset
detection accuracy at the expense of CPU time.
> Fixed a bug in the log_partition function which would sometimes cause
out of bounds errors
> Refactored and removed some functions that aren't needed anymore
Changes to sandbox.py:
> Sweeping changes to the visualization functions. The onset detection
functions are now combined after applying Gaussian blur to each onset
function individually. This improves the overall agreement between the
onset functions.
> Hyperbolic tan function is used to map saturated pixels to the range
[0, 1]
> Added a new rainbow generation function with more options than the old
one.
> Refactored most of the led update functions.
> The LED brightness is now being plotted instead of onsets
Led update functions no longer return a colorized array. Colorizing is
now done in led_visualization. It makes more sense for the led_update_X
functions to return 1D brightness arrays and then to apply color
afterwards. This should also improve performance somewhat. Also added a
new function leak_saturated_pixels() which allows saturated colors (>255
value) to leak into the adjacent color channels
